But what if a human driver would have seen that person while they were still 150 yards down the road? A human slows down ahead of time, monitors the situation, and doesn't have to react in the last 20 feet. It is very unlikely that the view of the human was totally obscured; though much MORE likely in the last 10 yards. When does the car start paying attention?
How far do these cars look ahead? In defensive driving, they teach you to look WAY up the road. 150 yards back from the intersection, you are more likely to see people running onto the road than 5 yards from the intersection; it may just be a flash of them seen between vehicles up ahead. Are these cars properly watching as they pull up? They should have to submit high definition video from the moment the car starts to when it stops, from the perspective of a driver. If that person that it hits becomes visible at any time and the AI doesn't show any reaction in some way, "she ran out in front of the car" isn't good enough if you're only paying attention 20 feet before the intersection. I know some people don't notice these things, but a lot of people do and it prevents accidents. I would rather have autonomous cars be modeled after defensive driving techniques and I am concerned that they are not.
- 10 million *carefully selected* miles with optimal conditions
- I asked an honest question if they are looking ahead at intersections. If they are gathering information about what is happening down the road then all my comments are moot. Companies don't seem to want to be open about this kind of information so assumptions are all we have to go by.
- Well, if there was a collection of experimental planes and one crashed, yes they would yank all the planes until they know what happened. Even now if a plane is found to have a serious flaw they do ground all planes that may have that flaw until it is checked. They certainly don't fly them over populated areas.
It's not speculation at all; it's an understanding of what successful human drivers do. It is a physical reality about the world that a car approaching that intersection needs to have an understanding of exactly what is going on in an intersection as it approaches. Information about the intersection may be 300, 200, 100 yards back and not available 5 yards away. And yes most people do do this, otherwise there would be a lot more human accidents. Every defensive driving course tells you to look around you. Some people stare at the car in front of them, agreed, but those are bad drivers. If AI is not capable of replicating all habits of good drivers then it is truly too early to have these things on the road. Despite the number of human accidents, the fact is that there are millions of cars on the road being driving by humans NOT getting in accidents and they are doing something right.
If an automated car is not gathering all information about an intersection from the visibility point on then AI is failing us, and will continue to kill people. Furthermore, people like you will continue to shrug their shoulders and wonder why.
Exactly. This is why I say automated cars aren't ready. If you can't see a hand while you are coming past a row of cars and not deduce that you should slow down, then AI will be killing people.
And a driver should watch what is happening as far down the road as possible. How far are automated cars watching down the road? Are they giving themselves time to slow down?
If this girl could have been seen 100, 200 yards back by a human, who would have slowed down by the time they reached the intersection, then the accident shouldn't have been a reaction call at all. How far ahead are these cars able to notice and track someone who is walking onto the road a block ahead? I know I see things like that all the time. And I might not see them again until it is too late because they go behind cars. But I would have slowed down and it wouldn't be an issue.
You're not understanding me. A human might well have seen the girl from 100 yards back but the AI might not have detected her until it was coming around the other stopped vehicles. A human would have 100 yards to slow down and be prepared to stop. If an AI is not looking that far ahead, it might end up with no time to stop.
It's more about observation than reaction time. So many times I've seen someone dart into the road a block and a half before the intersection. Sometimes you can't see all of them because there are so many cars around, but you see a jacket, or a head or an arm. Then you slow down and remove the requirement to react. I wonder how far ahead these cars are looking and if they are tracking everything moving as far as they can view? Or are they the driver that stares at the car ahead of them (ok in 360) and don't really understand what is happening way up the road.
Where I live, snow and ice gathers on the windshield. Some people get lazy in the morning and only scrape away a small section; just enough to see out of the windshield. We get tickets if we are seen operating a motor vehicle in this way. However, people still do it and can drive for quite a long time in this way without having issues. If we hit a person crossing the street while our vehicles were this way then we would surely be at fault. How do we tell whether sensors put on an automated vehicles are enough to consider that vehicle to be in a reasonable safe working condition? Or is automation just peeking through its hole in the windshield?
Well hopefully the determination is made how visible the pedestrian would have been to a human. If you are in a street and you see someone 200 yards down stepping off the curb, you slow down. I find it hard to believe that the situation in a wide road could be such that the pedestrian was totally obscured. If a human could see it then so should an automated vehicle.
I can't say I could imagine a car with such blind spots that you can't see a pedestrian you are about to hit in the road; that would require a fairly large section of whindshield to be blocked. A company would never be able to put such a vehicle on the road, so why are we allowing automated vehicles with such blind spots?
I think many people understand that you can't be fully attentive in a car that is driving itself. Therefore it is difficult to be in a position to take over effectively if you have to.
It has long been proven that humans get complacent behind automation and cannot take over effectively in all situations. Even if that's their job. It's just human nature and should not be used as a defense.
But if the sensors cannot e proven to work in all situations before getting into an accident, then that IS like impaired driving in a human. The safety driver is just a passenger being driven by a drunk in this case.
Slashdot Mirror
'Shadows' could also mean "Uber already payed me a shitload of money to downplay the situation".
But what if a human driver would have seen that person while they were still 150 yards down the road? A human slows down ahead of time, monitors the situation, and doesn't have to react in the last 20 feet. It is very unlikely that the view of the human was totally obscured; though much MORE likely in the last 10 yards. When does the car start paying attention?
But if that deer was in the middle of a wide open road at night, would you have missed it?
..and they will introduce five more bugs in the process, I'm sure.
How far do these cars look ahead? In defensive driving, they teach you to look WAY up the road. 150 yards back from the intersection, you are more likely to see people running onto the road than 5 yards from the intersection; it may just be a flash of them seen between vehicles up ahead. Are these cars properly watching as they pull up? They should have to submit high definition video from the moment the car starts to when it stops, from the perspective of a driver. If that person that it hits becomes visible at any time and the AI doesn't show any reaction in some way, "she ran out in front of the car" isn't good enough if you're only paying attention 20 feet before the intersection. I know some people don't notice these things, but a lot of people do and it prevents accidents. I would rather have autonomous cars be modeled after defensive driving techniques and I am concerned that they are not.
- 10 million *carefully selected* miles with optimal conditions
- I asked an honest question if they are looking ahead at intersections. If they are gathering information about what is happening down the road then all my comments are moot. Companies don't seem to want to be open about this kind of information so assumptions are all we have to go by.
- Well, if there was a collection of experimental planes and one crashed, yes they would yank all the planes until they know what happened. Even now if a plane is found to have a serious flaw they do ground all planes that may have that flaw until it is checked. They certainly don't fly them over populated areas.
It's not speculation at all; it's an understanding of what successful human drivers do. It is a physical reality about the world that a car approaching that intersection needs to have an understanding of exactly what is going on in an intersection as it approaches. Information about the intersection may be 300, 200, 100 yards back and not available 5 yards away. And yes most people do do this, otherwise there would be a lot more human accidents. Every defensive driving course tells you to look around you. Some people stare at the car in front of them, agreed, but those are bad drivers. If AI is not capable of replicating all habits of good drivers then it is truly too early to have these things on the road. Despite the number of human accidents, the fact is that there are millions of cars on the road being driving by humans NOT getting in accidents and they are doing something right.
If an automated car is not gathering all information about an intersection from the visibility point on then AI is failing us, and will continue to kill people. Furthermore, people like you will continue to shrug their shoulders and wonder why.
Exactly. This is why I say automated cars aren't ready. If you can't see a hand while you are coming past a row of cars and not deduce that you should slow down, then AI will be killing people.
And a driver should watch what is happening as far down the road as possible. How far are automated cars watching down the road? Are they giving themselves time to slow down?
If this girl could have been seen 100, 200 yards back by a human, who would have slowed down by the time they reached the intersection, then the accident shouldn't have been a reaction call at all. How far ahead are these cars able to notice and track someone who is walking onto the road a block ahead? I know I see things like that all the time. And I might not see them again until it is too late because they go behind cars. But I would have slowed down and it wouldn't be an issue.
You're not understanding me. A human might well have seen the girl from 100 yards back but the AI might not have detected her until it was coming around the other stopped vehicles. A human would have 100 yards to slow down and be prepared to stop. If an AI is not looking that far ahead, it might end up with no time to stop.
So we should just wait for a five year old to die needlessly?
It's more about observation than reaction time. So many times I've seen someone dart into the road a block and a half before the intersection. Sometimes you can't see all of them because there are so many cars around, but you see a jacket, or a head or an arm. Then you slow down and remove the requirement to react. I wonder how far ahead these cars are looking and if they are tracking everything moving as far as they can view? Or are they the driver that stares at the car ahead of them (ok in 360) and don't really understand what is happening way up the road.
Where I live, snow and ice gathers on the windshield. Some people get lazy in the morning and only scrape away a small section; just enough to see out of the windshield. We get tickets if we are seen operating a motor vehicle in this way. However, people still do it and can drive for quite a long time in this way without having issues. If we hit a person crossing the street while our vehicles were this way then we would surely be at fault. How do we tell whether sensors put on an automated vehicles are enough to consider that vehicle to be in a reasonable safe working condition? Or is automation just peeking through its hole in the windshield?
Well hopefully the determination is made how visible the pedestrian would have been to a human. If you are in a street and you see someone 200 yards down stepping off the curb, you slow down. I find it hard to believe that the situation in a wide road could be such that the pedestrian was totally obscured. If a human could see it then so should an automated vehicle.
Sure that would be great, but someone has to figure out how to get there without *killing people*.
Furthermore, were I live, driving with an obstructed windshield is a moving violation.
I can't say I could imagine a car with such blind spots that you can't see a pedestrian you are about to hit in the road; that would require a fairly large section of whindshield to be blocked. A company would never be able to put such a vehicle on the road, so why are we allowing automated vehicles with such blind spots?
I think many people understand that you can't be fully attentive in a car that is driving itself. Therefore it is difficult to be in a position to take over effectively if you have to.
But I thought automated cars have a split second reaction time, so that's not really a defense.
It has long been proven that humans get complacent behind automation and cannot take over effectively in all situations. Even if that's their job. It's just human nature and should not be used as a defense.
If the sensors are blind to humans crossing the road, then it is no different than a person driving impaired.
But if the sensors cannot e proven to work in all situations before getting into an accident, then that IS like impaired driving in a human. The safety driver is just a passenger being driven by a drunk in this case.
Ok... now just point to the part where it says it is legal to hit the person in violation of that.
But you can make your product safe.